“I've had a bunch of heads of R&D from pharma companies come by… I ask, what's your split, spending on your automation work cells versus your lab benches? And it's like 97% lab benches. Millions of square feet, filled to the brim with benchtop equipment, filled to the brim with lab technicians, all regulated space — and all running, get this, at best 40 hours a week. We have 168 hours in a week.”
Grow Everything Biotech Podcast · Alternative Proteins · September 2026
Kelly, Ginkgo Bioworks' CEO, is making the case for "lights out" autonomous labs, using an analogy between subways (fully automated, totally inflexible) and cars (manual, endlessly variable). His claim is that the money in pharma R&D is in the manual bench, not the robot cell, and that automating it would price experiments at the cost of reagents. Ginkgo announced the same day that it is building autonomous labs for MIT, Caltech, Northwestern and Maryland.
Transcript
Or not, though. Like you could control that too, but like, sure, it could, right? Okay. Now, here's the other thing, though. Miles traveled, cars and trucks versus subways. Any guess in the United States what the split is?
I'm sure it's probably like five and 95%.
99% cars and trucks. 99, 99 in human miles travel. Not because we don't know about trains, okay? But because we need variability, like we need it. Okay. All right. So now let us go to the lab. All right. Same graph, low variability, high amount of automation. We actually have subways in the lab. We call them automation work cells. And if you've ever seen, like if you're at a pharma company somewhere in a basement somewhere guarded by a bunch of automation people is an arm and it's moving plates through your high throughput screening system to hit a bunch of chemicals against cancer cells and see if you could pull a needle out of a haystack to discover a new drug. All right. Or if you've been to a diagnostic facility, you might have seen automated systems doing clinical diagnostics over and over again. Same test. It is great. It is fully automated. It can run overnight. No people, no mistakes, okay, that people make at the bench, but you got to want to do the same experiment that you did yesterday. It is totally inflexible. Low amount of automation, high amount of variability over here, the car. We have that. It's called the lab bench and the manual lab. And it is PhD scientists making up whatever experiment they want to do that day, grabbing reagents from the fridge, interacting with 50 to 100 pieces of benchtop lab equipment in a big room with 20 of their friends and all kind of doing the work of discovering drugs or finding a new trait for corn or whatever it might be that they're doing or academic work training like I would have done at MIT. Like that is our car. And here's the kicker. If you asked, and I've done this, we've had a bunch of head of RDs from pharma companies come by and visit Nebula, our big system here in Boston. I'm like, hey, what's your split? Spending on your work cells versus your lab benches. And it's like 97% lab benches. And just look around. There's millions of square feet for every major biopharma and biotech company. They have two to four million square feet of lab space. Filled to the brim with benchtop equipment, filled to the brim with lab technicians, all regulated space and all running, get this at best 40 hours a week. Okay, we have 168 hours in a week. I don't know if you've done the math, right? So like you're literally idling this insane infrastructure. I mean, like pharma companies spend, depending on which one we're talking about, one to three billion a year on not clinical trials, on those labs. And NIH spends 40 a year at all the universities on the same thing. And maybe they go, I used to work on some on the weekends. Like grad students work, maybe work a little bit more, but like definitely not working overnight.
Yeah. No, I would, I did pharma RD. I was at JNJ and GSK at a huge laboratory and there was like five of us. We had pretty much like a whole building and there was like 20 humans in there and there's so much space. There was very big laboratory equipment like the LCMS and things like that took up some space. But yeah, it's just like, I'm not going to work overnight. We're talking about nine to five.
Okay. So what we're, when I say autonomous lab, I'm not trying to replace the work cells, the subways at all. I think they're fine. Go buy them thermal cells and high-res sells them, bioServe, these like automation companies. Go buy them from wherever you buy them from. I'm trying to replace the lab benches. Okay. I believe that. That is your actual big spend and really where most of your drugs come from. But to get after that, I got to kind of do what Waymo did. I got to solve the variability. And that's a super hard problem. Like, that's the core of my technical problem. But again, I got to it honestly because at Gingo, we had been trying every combination of like semi-automated laboratories. Like our people plus lit walk-up liquid handling, plus our software on top of the walk-up liquid handling, all kinds of elaborate crap. And I'm telling you, I tried it all. It does not work well enough unless you get everyone out. You have to have lights out because the handoffs are too frictional. And so three years ago, we were like, all right, we got to go lights out. And we had acquired Zymergen, a Bay Area company who had their own robotics hardware. We made a new generation of that robotics hardware. And we've been working on the software nonstop. And we have all of our people able to use it here. They're like test pilots. And so the rate that we've been improving that place is pretty crazy. Like, if you come to see it now, it is, it's wild. Like, we have 105 robots, 160 devices, one big integrated system. It runs 24/7. Scientists submit random lab work to it every morning. It's not like we know ahead of time what they're going to submit. It can't do every protocol. We definitely run into all kinds of issues still. Like, it's not like we're done here, but like it's going to work. And that's wild. And that's back to my point about bringing my son to see that lab bench. Like, we got to close all those benches. And so we just announced today that we're building autonomous labs like the one I have here in Boston for MIT, Caltech, Northwestern, and University of Maryland through this new NSF White House Genesis mission program. So I'm pumped about it. I think this is like the way forward. But boy, have we tried like a million and one things that Gingko Automates make biology easier to engineer. It's still hard. It's still hard to engineer. So I like this is my bet over here. And then I think the guys making the big AI models, like the virtual self-stop is the other good bet right now. And those are the two new things being tried.
So what is your car then? Are they robotic arms or are they actual pieces of equipment? And I have seen videos, so I'm just doing this for our listeners.